Effects of introducing bundled payment and patients' choice of provider for elective hip and knee replacements in Stockholm county

Transcription

1 Effects of introducing bundled payment and patients' choice of provider for elective hip and knee replacements in Stockholm county Jonas Wohlin, Holger Stalberg, Oskar Ström, Ola Rolfson, Carl Willers, Mats Brommels February 2017

2 Karolinska Institutet Department of Learning, Informatics, Management and Ethics Tomtebodavägen 18 A, Stockholm ISBN:

3 Effects of introducing bundled payment and patients' choice of provider for elective hip and knee replacements in Stockholm county

4

5 Foreword By Professor Michael E. Porter As countries around the world grapple with unsustainable costs and erratic quality of health care systems, a central theme is emerging. The only real solution is to substantially improve the value of care, where value is the patient health outcomes achieved relative to money spent. There is an urgent need for bold steps to move from volume-based health care delivery to value-based health care delivery. As introduced in my 2006 book with Elizabeth Teisberg, Redefining Health Care, this will require a shift to multidisciplinary, team-based care organized around patients medical conditions, to measuring the patient outcomes that matter for each condition, and the overall costs of delivering those outcomes. Over the last several years, health care providers globally have begun to embrace these principles in numerous medical fields. In moving to a value-based system, a key enabler is shifting to bundled reimbursement, or reimbursement that covers the full cycle of care for the patient s medical condition. Fee for service rewards volume of services, not value. Global capitation, or a single payment covering any medical need that arises, exposes providers to risks that they cannot control, has been prone to encourage limitation of services only, and lacks accountability for specific outcomes. Bundled reimbursement aligns payment with value in delivering care for the patient s particular medical need, and aligns payment with the outcomes the provider team can control. Coupled with rigorous outcome measurement, then, bundled reimbursement encourages and rewards relentless value improvement. The benefits of bundled reimbursement are now becoming widely recognized, but implementation has proceeded slowly. Bundled reimbursement compensates the entire care team rather than today s siloed reimbursement for individual services. It also requires understanding the true costs of delivering care over the care cycle, something providers today rarely understand. Bundled reimbursement also works best when combined with rigorous outcome measurement, which is still in its infancy in most countries. There is a rapidly growing number of bundled pricing pilots and initiatives, but large-scale implementation is still in the early stage. 5

6 Foreword By Professor Michael E. Porter One of pioneer examples is the Vårdval höft-och knäprotesoperationer (Choice of Care in Hip and Knee Replacements) in the County of Stockholm, Sweden. Since January 2009, all non-complex hip and knee replacements in the county of 2 million people have been reimbursed with a bundled payment, covering over 4,000 procedures per year. The bundle includes all services from pre-operative evaluation to outpatient follow-up, and incorporates care guarantees covering infection and revision surgery. In partnership with the wellestablished Swedish Hip Arthroplasty Register, reporting of outcomes is mandatory. The results of this pathbreaking new reimbursement model have been striking, as this report demonstrates. Not only has the new reimbursement model encouraged providers to make numerous improvements in care, but complications have dropped substantially and waiting time has disappeared. The County Council is refining the reimbursement model and has extended it to other conditions. This case illustrates that bundled reimbursement is not only feasible, but value enhancing. Stockholm s leadership provides an important benchmark to guide the design and implementation of bundled reimbursement globally. Sincerely, Michael E. Porter, PhD, MBA Bishop William Lawrence University Professor Director, Institute for Strategy and Competitiveness Harvard Business School 6

7 Executive summary BACKGROUND Previous research has shown that transparency around health outcomes and changes to reimbursement influence how healthcare is delivered. It has been suggested that bundled payment, where providers receive a package price for the entire cycle of care, and where financial risk for complications is transferred to the provider, enables providers to work differently with potentially positive effects on outcomes and cost. In 2009, the Stockholm County Council (SLL) introduced bundled payment for primary hip and knee replacements and also allowed citizens to choose freely among accredited providers ( Patient choice program ). Compared to other countries, Sweden performed well even before the reform in terms of complications and prosthesis survival. However, waiting times were unacceptably long and many patients had to wait for more than one year for surgery. In 2010, the Karolinska Institutet was commissioned by the SLL to analyze the effects of the reform from a patient value perspective. The study has been performed in collaboration with the Institute for Strategy and Competitiveness at Harvard Business School, and the Swedish Hip Arthroplasty Registry. RESULTS The combined introduction of bundled payment and adoption of patients choice of provider resulted in higher production capacity and reduced waiting times. There were, however, no signs of patients being operated on earlier in the progression of the disease than what was medically justified or that patients would undergo surgery on other indications than previously. A number of providers implemented changes in order to reduce the cost of treatment as well as the risk of complications. Changes included standardization of the treatment process, development of manuals and checklists for staff, certification of personnel, introduction of financial incentives for staff (tied to low rates of complications), and the introduction of additional follow-up visits to identify complications at an early stage. Private providers 7

8 Executive summary implemented changes to a greater extent than public providers. Providers reported an overall level of satisfaction with the new model. Quantitative analysis shows that the risk of reoperation after primary hip/knee replacement in Stockholm dropped by 26 percent after the introduction of bundled payment and patient choice of provider (2010 compared to ). This risk reduction can be partially explained by a general reduction in risk among healthcare providers who treated Stockholm patients before, and partially by a shift in volume to providers with lower risk. A difference-in-difference analysis was carried out to investigate whether these results were in fact associated with introducing the new model, or whether they seemed to depend on secular trends applicable to regions outside of Stockholm (new surgical methods, quality initiatives, more resources, quality registry monitoring etc.). Results from the analysis showed that improvement in complication rates observed in SLL was not caused by trends present in the seven comparison regions. If anything, the interaction model indicates that the SLL results were even more favorable than what was found when studying SLL alone. Private providers reduced their risk of complications more than public providers. This is consistent with the results from the interview-based analysis in which private providers reported taking greater action to improve the treatment process than public providers. No change was observed in patient-reported outcome measures (PROM). Analysis of cost at provider level showed a reduction in costs over the full cycle of care of approximately SEK 11,300 per patient (14 percent). At the payer level, reduced prices and risk transfer led to a cost reduction per treated patient of SEK 16,500 (20 percent) for SLL, corresponding to a total of SEK 60 million per year. For the county, the total cost of primary hip and knee replacement surgery fell by 3 percent, even though total volumes increased by 21 percent. Sick leave during the year before and the year after primary surgery declined in total by 17 percent (38 net days of sick leave). A corresponding reduction was observed across the rest of the country, and the change observed in SLL was probably due to a combination of reduced queues and political reforms. CONCLUSION AND RECOMMENDATIONS After the reform, waiting times for primary hip/knee replacements decreased, and costs over the full cycle of care, as well as the risk of complications, were significantly reduced. Providers were also generally satisfied with the bundled payment model. Several recommendations have been set out at the end of this report. Amongst these recommendations, the most important ones are to secure financing for education and to expand the scope of the reform to also include ASA 3 patients (currently ASA 1-2). 8

9 Contents TABLES...11 FIGURES...13 INTRODUCTION Background The new model and hip/knee replacement surgery in Stockholm Purpose, limitations and outline DATA AND METHODS...19 RESULTS Effects on production and accessibility Effects on providers ways of working and care process Effects on health outcomes Effects on resource consumption and costs Effects on the patients experience of care Healthcare providers view on the new model and its effects CONCLUSIONS AND DISCUSSION...41 AUTHORS RECOMMENDATIONS REFERENCES APPENDIX Differences in analyses compared to the Swedish Hip Replacement Register A Summary Definitions and additional results

10

11 Tables Table 1. Complications demanding inpatient stay after primary hip/knee replacement...30 Table 2. Average number of net sick-leave days one year before and after primary surgery, in Stockholm and the rest of Sweden...34 Table 3. Resource utilization before and after introduction of the new model Table A1. Study population patient characteristics (primary hip/knee replacement surgeries )...50 Table A2. Cox model on the risk of hospitalization due to prosthetic related complications diagnosed within 2 years of the index operation Table A3. Cox model on the risk of hospitalization with prosthetic related reoperations within 2 years of the index operation Table A4. Cox model on the risk of prosthetic replacement surgery within 2 years of the index operation Table A5. Logistic mixed-effects model on the risk of hospitalization due to cardiovascular events within 30 days of the index operation Table A6. Number of patients with at least one hospitalization due to prostheticrelated potentially avoidable adverse events within 2 years ( )...55 Table A7. Procedure codes used to identify reoperations (including revision/ removal) and cardiovascular events Table A8. Number of patients with reoperations ( ) Table A9. Number of patients with replacement or removal...57 Table A10. Total number of primary hip/knee replacement surgeries Table A11. Total production of primary hip/knee replacement surgeries in 2008 vs and in 2008 vs Table A12. Total production of primary hip/knee replacement surgeries under the new model in 2009 and

12 Tables Table A13. The effect of various factors on the risk of reoperation within 2 years of the index operation Table A14. Use of cement-free prosthetic or hybrid technology among various age groups under the new model during (only hip replacement)...58 Table A15. Pharmaceutical use per patient and on aggregate level

13 Figures Figure 1. Production volumes and proportion of patients with waiting times for surgery >90 days Figure 2. Production volumes 2008, 2009, Figure 3. Average length of stay per year...26 Figure 4. Average length of stay per provider...27 Figure 5. Average length of stay per provider for patients covered by the new model Figure 6. Adjusted relative length of stay for all patients, with confidence intervals (95%)...28 Figure 7. Proportion of patients undergoing reoperation within two years after primary hip/knee replacement...32 Figure 8. Sick-leave patterns for Stockholm patients with at least one net day of sick leave the first year after primary surgery...33 Figure 9. Production volume and total payer cost Figure 10. Payer cost per patient...37 Figure A11. Cumulative unadjusted risk for reoperation prior to and following the reform (95 % confidence interval) Figure A12. Charlson comorbidity index by type of healthcare provider Figure A13. Charlson comorbidity index by healthcare provider ( )

14

15 1 Introduction 1.1. BACKGROUND The overall goal of each and every healthcare system is to create as much good health as possible for its population, with the limited available resources. Within healthcare, however, variation in methodology as well as results, in terms of quality and cost, between countries, regions and producers is larger than in the majority of other sectors. At the same time, in many countries, healthcare costs have increased faster than GDP, which is not a development that is sustainable in the long term [i]. It has been suggested that one possible reason for this variation and unsustainable cost increases, in addition to new ways of working and an aging population, is that healthcare providers, in accordance with tradition, are organized according to medical specialties and not according to diseases and treatment. Additionally, healthcare management and governance (such as monitoring and reimbursement) are often focused on the individual activities that healthcare professionals perform instead of the treatment results they achieve. This way of organizing healthcare has created difficulties for healthcare providers in coordinating and continuously developing their treatment process to achieve the best possible health outcomes with as few resources as possible [ii]. Bundled payment. In both literature and in practice a new type of reimbursement model is being developed that has been theoretically well placed to contribute to the developments outlined above. Within this model, healthcare providers are reimbursed with a package price for the whole or the greater part of the continuum of care. This type of compensation is known internationally as bundled payment. Principles underlying bundled payment can be divided into three main components: package price, performance compensation and individual adjustment. Package price. The basic principle is that the provider receives a package price for an entire chain of care instead of being reimbursed for each individual activity. The package price may also include compensation for expected costs related to potentially avoidable complication and thus include a warranty reimbursement, which results in the provider 15

16 Introduction bearing financial responsibility for these predefined complications. By compensating providers with a package price, parts of the financial risk shifts from the payer to the provider, which means that the provider is given both more freedom and more responsibility to ensure effective and high-quality patient flows. Performance compensation. Considering that parts of the reimbursement are based on agreed upon performance indicators (health outcomes or other key figures), which are relevant to the specific patient group, the provider can invest in activities that are expected to improve their performance. Individual adjustment. Some bundled payment systems are designed in such a way that the package price and performance compensation are adjusted according to treated patients individual conditions (disease-specific factors, comorbidities, demographics, etc.). This means that patients with different conditions can be treated under the same reimbursement model. Research has previously shown that healthcare management and governance have an impact on how care is delivered [iii, iv]. Hypotheses suggest that, by increasing the degree of freedom and responsibility shifted from payers to providers, bundled payment can enable and encourage healthcare providers to be innovative and change their practices in order to improve quality and reduce resource utilization [4, v, vi] THE NEW MODEL AND HIP/KNEE REPLACEMENT SURGERY IN STOCKHOLM In January 2009, the healthcare administration at Stockholm County Council (SLL) introduced, in accordance with the Act on System of Choice (LOV), a so-called patient choice for primary elective hip and knee replacement surgery, and at the same time introduced a bundled payment model for the patients covered by the patient choice. The new model required healthcare providers who wished to perform these operations to undergo an authorization. Patients were free to select among authorized providers, which were not limited in production capacity. Authorization criteria included, among other requirements, reporting of quality data, that the operating surgeon had to perform at least 50 operations per year, and that the operating room had to meet certain air quality requirements. The new model was only available for patients without comorbidities that caused functional limitations (ASA 1-2), totaling approximately 78 percent of all hip and knee replacement patients with total replacement in Fractures, as well as fracture repair failures, were not covered by the new model. The bundled payment model was introduced with the following design: Package price: Healthcare providers received a package price of SEK 56,300 for the continuum of care, including diagnostics, surgery with follow-up care, prosthetic costs, 16

17 Introduction and the necessary pre-surgical and post-surgical visits. To ensure that the initial diagnosis took place in primary care (and that the cost of the medical work-up was born by the primary care unit), compensation for the initial visit was reduced by 50 percent along with the introduction of a fine for providers who had more than 35 percent of initial visits that did not lead to surgery. Moreover, there was no specific remuneration for radiological examinations. Providers also had to assume the responsibility for any complications and became financially responsible for potentially avoidable adverse events (PAAE) that were related to the primary surgery and which occurred up to two years after the primary surgery (five years in cases where infection treated with antibiotics occurred during the first two years). Performance compensation. In respect of compensation, 3.2 percent was withheld and paid out as performance compensation if providers reached certain predetermined targets (mostly process measures). Individual adjustment. Whereas the patient group included in the new model was considered to be homogeneous, individual adjustment was not deemed to be necessary. Providers treatment outcomes would be followed by SLL s patient administrative system as well as through the national quality register. Patient satisfaction with the treatment process was measured using a specially designed questionnaire that the patient received after completing treatment. Previously, the majority of these operations were performed in emergency hospitals and reimbursed according to the NordDRG model. Additionally, specific volumes were procured under the Public Procurement Act (LOU) by SLL and sometimes by public healthcare providers who used private healthcare providers as subcontractors. In comparison with other countries, Sweden performed favorably, with as good or better prosthetic survival rates for hip replacements [vii, viii]. Still, a large proportion of patients in the SLL had to wait more than 90 days for surgery [ix] PURPOSE, LIMITATIONS AND OUTLINE In 2010, the Stockholm County Council initiated a joint study together with the Karolinska Institute, which was expanded to include collaboration with the Harvard Business School and the Swedish Hip Replacement Register. The aim was to study how the adoption of the new model affected healthcare providers professionally and whether it had value-creating effects for the health system in terms of improved medical outcomes and/or lower costs. In addition, recommendations would be given for further development. It is important to note that the introduction of the new model included two components, each of which, according to different logics, may have influenced the added value for the healthcare system: firstly, the patients choice of provider, and secondly, the bundled payment model. 17

18 Introduction On March 2, 2012, a preliminary report was published by the SLL, describing the early findings, conclusions and subsequent recommendations. In 2012 and 2013, the majority of the research project s quantitative analysis was conducted, which serves as the basis for this expanded report. In 2015, an additional analysis was carried out to investigate changes in complication rates in SLL in comparison with other counties. The report includes an appendix that contains in-depth information pertaining to the results and methods of analysis, including comparability with quality register data on complications. 18

19 2 Data and methods This report is based on interviews as well as an analysis of patient-level and aggregate-level register data. Both the analysis plan and the data collection plan have been approved by Stockholm s regional ethical review board (2011/5:6). In addition, the Sveus research database was utilized for in-depth analysis of the trends in complication rates in SLL compared to other counties affiliated with Sveus ( [x]. This took place under a separate research approval from Stockholm s regional ethical review board (2011/ and supplement 2013/ ). The qualitative analysis is based on interviews with representatives from 10 healthcare providers. All interviews were recorded and conducted using a semi-structured approach, that is, by initially posing broad questions about the new model, which were later complemented by specific questions on predefined topics. The material was transcribed and subject to content analysis. Representatives of SLL were interviewed during the spring of 2010 and an ongoing dialogue has taken place between SLL and the research team throughout the project period. The interviews with SLL have not been subject to content analysis. The quantitative analysis is based on linking information from SLL s administrative health database (GVR/VAL), Swedish Hip Replacement Register (SHPR), Prescribed Drug Register (PDR), sickness absence data from the National Social Insurance Agency (FK), and socioeconomic and demographic variables from Statistics Sweden (SCB). Additionally, an incomplete database was obtained from the healthcare providers surgery planning system with the ASA classification during the period Aggregated data on waiting times and surgical volumes was obtained from the SLL, Swedish Hip Replacement Register (SHPR) and the Swedish Knee Arthroplasty Register (SKAR). The research database includes 16,177 patients, registered as residents of Stockholm County, who underwent primary hip or knee replacement surgery during within the SLL. Patients who had surgery for tumors or fractures were excluded. All analyzed surgeries had at least 12 months follow-up time, and 81 percent had 24 months of followup. The primary hip or knee replacement surgery that was analyzed for each individual is designated in this report as the index operation. 19

20 Data and methods To understand the impact of the new model on the patient group as a whole to the greatest extent possible, all patients who underwent surgery during the period were included, comprising both patients covered by the new model and those covered by DRG reimbursement. Wherever possible, statistical analyses were adjusted to account for differences in patient characteristics, between groups and over time, hereinafter referred to as the case mix. Analyses were adjusted for age, gender, somatic comorbidity, depression diagnosis, educational level, country of birth, disposable income, prior primary joint replacement in another joint, and type of prosthesis (see overview in tables A1 A4). Statistically significant differences were assessed at the 5 percent level, in accordance with scientific practice. The results are reported in the body of the report, with associated figures and tables located in the Appendix. The surgeries included in the material under study were carried out by 10 different healthcare providers, of which five were owned by SLL ( public providers ) and five were privately run ( private providers ). This report also utilizes a division into specialist centers and emergency hospitals. In this report, a specialist center is a unit that performs hip and knee replacement surgery, and which is not an emergency hospital. Capio St. Göran s Hospital is the only emergency hospital that is run privately. The remaining four of five private providers were considered specialist centers. Below is a description of the method for identifying complications or potentially avoidable adverse events (PAAE): Orthopedic complications or PAAE were defined by at least one diagnostic code for complications (Table A6) or procedure code for reoperation (Table A7) recorded in connection with an episode of inpatient care within two years of the primary surgery. The codes were based on the coding guide for complication diagnoses from the Swedish Orthopedic Association. Since multiple inpatient care episodes, sometimes with different diagnoses/ procedures, could often have one single, original cause, repeated complications diagnoses/ procedures were counted as only one complication. Complication codes were also identified during the inpatient stay during which the primary surgery was performed. Identified orthopedic complications were handled hierarchically, with prosthetic removal/replacement being counted prior to reoperation, and reoperation prior to inpatient stay with a sole diagnosis of orthopedic complication. For example, a patient who first had an inpatient visit for an infection that had been treated with debridement, but who subsequently underwent a prosthetic removal/replacement, would only be defined as a prosthetic removal/replacement in this analysis. Orthopedic complications were estimated through survival analysis, taking into account that patients may have different lengths of follow-up following primary surgery. Follow-up time of an individual patient ended if the patient suffered a complication, had a new primary operation in another joint, died, or moved to another county. Cardiovascular events (myocardial infarction, deep vein thrombosis, and pulmonary embolism) were identified by relevant diagnosis codes (Table A7) registered during inpatient visits commencing within 30 days of the primary surgery. 20

21 3 Results This section describes the results emerging from effects observed across six dimensions: production and accessibility (3.1), healthcare providers practices and care process (3.2), health outcomes (3.3), resource utilization and cost (3.4), patients experience of care (3.5), and the providers views on the new model and its effects (3.6) EFFECTS ON PRODUCTION AND ACCESSIBILITY A principal reason for introducing the patients choice of provider was to reduce the long queues by increasing accessibility. At the same time, concerns existed within SLL that the patients choice of provider would lead to patients undergoing surgery earlier during the course of the disease than medically motivated, or to patients having surgery on the basis of other indications than before. There was also a concern that sicker patients who were not included in the new model (ASA 3-4 patients) would have poorer access to healthcare. These aspects are analyzed in this section. Temporarily increased volumes. When the new model was introduced in 2009, the total volumes of elective hip and knee replacement surgery in Stockholm increased by 20 percent (care choice and DRG-funded), which then decreased by a total of 4 percent until The decrease was driven by the patient-choice funded portion of the surgeries, which decreased by 12 percent during the same period. The DRG-reimbursed surgeries, however, increased by 20 percent during the period (Table A10). No queues for patients covered by the new model. The overall proportion of patients who waited more than 90 days for primary hip or knee replacement surgery (the new model and DRG-reimbursed) fell between 2008 and 2010 by 23 percentage points (Figure 1). Since healthcare providers reported that no queues existed for patients covered by the new model, it is likely that those patients who were reported as having waited more than 90 days after the introduction of the new model (including for medical reasons and self-selected waiting) consisted primarily of patients waiting for DRG-reimbursed surgeries. 21

22 Results Two providers reported that they have successfully begun to allow patients to book their own surgery time during the appointment for therapy decisions. Surgery volumes 37% Patient share with waiting time >90 days % 35% 30% % % 13% 10% 5% % 15% 10% 5% DRG Vårdval New Andel Patient väntande share waiting model 0% Figure 1. Production volumes and proportion of patients with waiting times for surgery >90 days. Includes all primary elective hip/knee replacements in Stockholm Proportion of patients with waiting times >90 days include both medically justified waiting and self-selected waiting. Increased equal access to care. To understand whether introducing the new model affected access to care for specific groups, changes in sociodemographic variables among patients undergoing surgery were studied (educational level, marital status, country of origin and disposable income). In addition, preoperative pain level has been analyzed for different groups. The analysis shows that the increase in accessibility occurred alongside an equal rise in volumes for all groups. Detailed examination showed that patients born outside the EU-27 reported a higher preoperative level of pain (p<0.05) than patients born in Sweden. Such a disparity can possibly be explained by cultural differences in the way pain is perceived and presented [xi, xii]. The difference between patients born outside the EU-27 and the other groups increased slightly after the introduction of the new model, however not statistically significant (p=0.23). More surgeries took place at specialist centers and fewer at emergency hospitals. All eight healthcare providers who previously operated on publicly financed Stockholm patients as well as two additional private specialist centers were authorized during the introduction of patients choice of provider. From 2008 to 2009, shifts took place in production from public 22

23 Results providers (-18 percent) to private providers (+107 percent), and from emergency hospitals (-13 percent) to smaller specialist centers (+218 percent) (Figure 2). EmergencyAkutsjukhus hospitals Public Offentlig Specialistcentra Specialistcenter centers Private Figure 2. Production volumes 2008, 2009, Includes all primary elective hip/knee replacements in Stockholm during 2008, 2009 and Emergency hospitals increasingly took care of patients with high comorbidity. When analyzing the Charlson Comorbidity Index (CCI), a well-established comorbidity index [xiii], it was found that emergency hospitals increasingly cared for patients with high comorbidity, while specialist centers took care of patients with less comorbidity (p<0.05). On average, CCI increased from to (+14%) for emergency hospitals, and decreased from to (-15%) for specialist centers. The new model does not appear to have led to operating based on other indications. As the adoption of the new model meant that past constraints on volume production were removed, and since the volume also initially increased, it is relevant to study whether patients with less discomfort underwent surgery. Calculations of average CCI showed a total reduction of approximately 7% after introduction of the new model. One possible reason for this could be that those patients who, prior to the adoption of the new model, were in line and therefore had not been prioritized for surgery, had, on average, slightly lower comorbidity than those who had undergone surgery. It was found that the adoption of the new model had 23

24 Results not prompted any change in the mean age at time of surgery. Quantitative analysis shows that preoperative pain levels and patient-reported quality of life for patients who underwent hip replacement surgery remained unchanged or deteriorated slightly. All in all, this indicates that patients did not appear to have received surgery earlier in the course of the disease than what was medically justified. This finding is consistent with that of SLL from an audit that was performed and which included extensive review of medical documentation. Thus, initial volume increase was probably related to a need that had accumulated during the time preceding the introduction of the new model EFFECTS ON PROVIDERS WAYS OF WORKING AND CARE PROCESS The design of the reimbursement system aims to enable and stimulate healthcare providers to be innovative and to provide healthcare in a way that accomplishes the best possible health outcomes at the lowest possible cost. This section analyzes the impact on healthcare providers work based on interviews with organizational representatives in combination with quantitative analysis of register data. Fundamental changes in the care process. Five of the nine providers who performed these surgeries prior to the adoption of the new model stated that the new model design resulted in them making changes to the treatment process or their way of working in order to enhance productivity and/or to reduce the risk of complications. Three of these providers, all private, reported that they implemented major change projects such as study visits to other providers in Sweden and abroad, process mapping, development of manuals and checklists, development of information for patients, staff training and certification. Three providers reported that they introduced additional follow-up visits with the purpose of having trained nurses remove surgical stitches and at the same time identify any infections at an early stage. Previously, the stiches were removed within primary care. And then one should not hide the fact that this warranty also brings forward an awareness. One feels that it costs to have complications. I must say, it increases quality. Healthcare provider during an interview Increased number of surgeries per room, team and day. Three healthcare providers reported that they increased the number of surgeries per room, team and day from three to between four and five, as a consequence of the adoption of the new model. All three providers accomplished this by reducing switching time and in no cases should this have resulted in longer working hours for staff. In addition, a redistribution of surgical volumes took place to the benefit of providers with higher productivity in the operating room. An estimate of resource utilization based on interviews with healthcare providers in combination with the observed changes in the distribution of operation volumes presented in Figure 2 shows 24

25 Results that the average number of surgeries per room, team and day increased from approximately 3.1 to 3.6 from 2008 to This corresponds to an increase in the utilization rate of fixed surgical resources by approximately 16 percent. For patients covered by the new model, approximately 3.7 surgeries were performed per room, team and day during We have also increased productivity in the operating room by thirty-three to fifty percent. Healthcare provider during an interview Reduced length of inpatient stay after surgery, but with major differences between healthcare providers. Duration of hospital stay in surgical departments in connection with the surgery has, as similar to the rest of the country, continuously decreased during the last decade. When adjusting for trends over time, gender and type of surgery, a significant reduction in length of inpatient stay after introduction of the new model was noticed. By contrast, after 2010 the continuous reduction that was observed during the period before the adoption of the new model ceased, and the length of stay stabilized at around four days during The absolute reduction in length of stay was even greater when postoperative stay at geriatric and/or rehabilitation department was included in the analysis. The reduction then amounted to a about one day (crude values): from approximately 6.7 days in 2008 to 5.8 days in 2009 (Figure 3). A large part of the reduction appeared to be due to the shift of surgical volumes from emergency hospitals with longer average inpatient stay to specialist centers with shorter average inpatient stay. The qualitative analysis was consistent with the quantitative regarding the reduction in length of stay: four providers reported that they reduced treatment time by an average of one day and that increasingly patients were admitted the same day as the surgery was performed instead of the day before the procedure. After the adoption of the new model, providers were also responsible for rehabilitation costs incurred at external rehabilitation units. A number of providers reported that they had therefore actively stopped sending patients to external rehabilitation units following the adoption of the new model, as they argued that patients sometimes remained for unjustifiably long periods of time in these units, which the providers did not want to pay for. In the quantitative analysis, it appeared that the use of external geriatric and rehabilitation units decreased (crude values) from 1.42 days in 2008 to 1.10 in 2009 and 0.92 in Initial analysis of treatment time led on to an in-depth analysis of differences between healthcare providers. Figure 4 and Figure 5 show the unadjusted average length of stay per provider and Figure 6 shows the relative differences in length of stay between providers after adjusting for patient characteristics available in the data. There were significant differences between different healthcare providers with a clear line of division between emergency hospitals and specialist centers. 25

26 Results Length of stay, care days PUOH PAAE Post-opgeriatrik geriatrics Post-op rehab Operationstillfället Surgery Specialistkliniker centers Emergency Sjukhus hospitals Totalt Figure 3. Average length of stay per year. All primary elective hip/knee replacements Inpatient stay for potentially avoidable adverse events (e.g. infection or reoperation) related to the primary surgery is separated from inpatient stay related to the surgery, geriatric care and rehabilitation. Length of stay split by specialist center and emergency hospital, respectively. 26

27 Results Emergency hospitals Specialistcentra centers Figure 4. Average length of stay per provider. All primary elective hip/knee replacements Akutsjukhus Emergency hospitals Specialistcentra Specialistcenter centers Figure 5. Average length of stay per provider for patients covered by the new model. Primary elective hip/knee replacements under the new model

28 Results 2,2 Adjusted Justerad relative relativ length vårdtid of stay 2,0 1,8 1,6 1,4 1,2 1,0 Figure 6. Adjusted relative length of stay for all patients, with confidence intervals (95%). Relationship between providers length of stay (including geriatric care and rehabilitation) for all patients, adjusted for age, gender, depression diagnosis, education level, country of birth, disposable income, prior primary joint prosthesis, co-morbidity and prosthesis type. Ortopediska huset = 1.0; value of 2.0 corresponds to 100% longer average length of stay compared to Ortopediska huset, adjusted for patient characteristics as mentioned above. No significant changes in surgical method or choice of prostheses. No healthcare providers suggest that they have changed surgical technique or type of prosthesis as a consequence of introducing the new model. Quantitative analysis shows that the hips that were operated on without cement and with hybrid technology increased during the observation period (after 2007), but that there have been no significant changes following the adoption of the new model. There are however differences in the choice of hip prosthetic between providers (see Table A14). A private healthcare provider reported having negotiated prices down for special prosthetics by about 30 percent after the adoption of the new model. No other providers reported changes in prosthetic prices and no providers reported changes in prosthetic supplier resulting from the adoption of the new model. Reduced prescriptions of opioids and antibiotics. Analysis of data from the Prescribed Drug Register shows that the prescription of opioids decreased by 21 percent. The prescription of other analgesics increased by 7 percent and the prescription of antibiotics declined by 8 percent after the adoption of the new model (see Table A15). 28

29 Results Minor changes in facilities/equipment. One specialist center reported investing in improved air quality (laminar flow) in the operating room resulting from the requirements within the new model. No other changes to facilities or equipment were reported. Private providers introduced financial incentives for staff. One private provider reported that a small incentive for staff based on volume was introduced, but that it has not resulted in any positive effects. Two additional private providers planned to introduce their own respective financial incentives for staff: one based on patient satisfaction and shortterm results and the other based on cost savings related to avoided complications and the complications warranty. Public providers expressed that it was not possible for them to introduce financial incentives for staff. Marketing through the improvement of websites and information for other healthcare providers. Four private providers and one public provider reported that they actively worked with patient recruitment. Their methods included improved websites, letters to general practitioners, as well as information meetings for general practitioners and physiotherapists. No providers reported marketing their services directly to patients, aside from providing improved information to those patients who sought out their services on their own. Providers reported that an overwhelming majority of patients chose the provider that their general practitioner had recommended. Improved service through increased accessibility and a better structured care process. Five providers were of the opinion that the level of service provided to patients increased after the adoption of the new model. Increased accessibility over the phone, improved structure of the continuum of care, and better patient information, as well as the ability to select date for surgery at the time of therapy decision were cited as primary reasons. One cannot (or should not) expect to benefit without making some effort, rather we actually have to work for it. We must create a good reputation around town. It is essential for us to continue to have our jobs and being able to do these things. Healthcare provider in an interview 29

30 Results 3.3. EFFECTS ON HEALTH OUTCOMES The analyses in this section aimed to investigate whether the health outcomes, i.e. the treatment result, changed after the adoption of the new model. In this report, health outcomes are divided into complications or PAAE registered in SLL s administrative healthcare database (GVR/VAL), and patient-reported health outcomes (PROM) reported to the Swedish Hip Replacement Register (SHPR) for patients who underwent hip replacement surgery. Section 5 presents the analysis of sick leave, which can also be seen as a health outcome. Reduced risk of complications. Four categories of complications or PAAE have been analyzed (definitions are presented in Tables A6-A7): Inpatient care episodes with complications diagnosed, within two years of primary surgery Inpatient reoperation within two years of primary surgery Inpatient prosthetic removal/replacement within two years of primary surgery Cardiovascular event within 30 days of primary surgery The risk of complications decreased across all categories following the adoption of the new model (Table 1). The degree of orthopedic complications further decreased when comparing surgeries carried out in 2010 with the 2007/2008 population. Table 1. Complications demanding inpatient stay after primary hip/knee replacement. Results adjusted for simultaneous changes in patient characteristics (age, gender, depression diagnosis, educational level, country of birth, disposable income, prior primary joint prosthesis, comorbidity and prosthesis type). Complications Risk 2007/2008 Change 2009/2010 Statistical significance of change Change 2010 Statistical significance of change Orthopedic complication within 2 years 6% -18% p= % p=0.00 Reoperation within 2 years 5% -23% p= % p=0.00 Revision/removal within 2 years 2% -19% p= % p=0.01 Cardiovascular event within 30 days 1% -44% p=0.01 According to the analysis performed, there were no significant differences in the risk of complications between hospitals and specialist centers prior to the adoption of the new model. However, private providers decreased their risk of complications more than public providers after the adoption of the new model. The total risk reduction can partly be explained by a general risk reduction among healthcare providers who treated SLL patients prior to the adoption of the new model, and partly by volume shifts to providers with a lower level of risk. This is consistent with the results of the qualitative, interview-based analysis, in which 30

31 Results private providers reported to a greater extent than public providers that they implemented changes to the care process with the aim of reducing the risk of complications 1. The analysis also shows that the patients sociodemographic and medical conditions prior to the surgery had a significant effect on the risk of complications. Patients with a disposable family income in the fourth quartile (second highest) were at a 23 percent lower risk compared to those who were in the first quartile (lowest family income). Similar effects could be identified for other categories of complications (see Tables A2-A5). It should be noted that the follow-up time in this analysis was limited to two years and that analysis of the change over a longer period was not possible within the scope of this study. However, in the future, this analysis should be supplemented with analyses of complications occurring within 5 10 years. It should also be noted that the method used in this report to detect complications is not directly comparable to the numbers presented in the annual reports of the Swedish Hip and Knee Replacement Registers; for a full description of the differences see the summary in the Appendix. Significant reduction in the frequency of orthopedic complications persisted when compared with trends in other counties. Difference-in-differences analyses were carried out to investigate whether the reduction in orthopedic complications was actually associated with introduction of the new model, or whether it seemed to depend on secular trends applicable to regions outside of Stockholm (new surgical methods, quality initiatives, more resources, quality register monitoring etc.). Results from the analysis showed that improvement in complication rates observed in SLL was not caused by trends present in the six comparison regions. If anything, the interaction model indicates that the SLL results were even more favorable than what was found when studying SLL alone. Difference-in-differences analyses were conducted using the Sveus research database (see section 1.2). Changes in the rate of inpatient reoperations are presented in Figure 7. 1 The risk for complications was significantly lower in private providers compared to public providers. In a report from the The Swedish Agency for Health and Care Services Analysis (MYVA) differences were highlighted between various types of hospitals [1]. However, this analysis had somewhat other prerequisites: i) results in this report are based on analyses of patient administrative data directly from the healthcare providers medical records whereas the MYVA results are based on quality registry data, manually filled out and separately reported by the provider ii) somewhat differing definitions of outcomes, for details see Appendix iii) within this report results were adjusted for simultaneous changes in case mix factors stated above whereas the MYVA report included adjustments only for age, gender, BMI and ASA iv) MYVA s comparison treated Sweden as a whole v) differences in time horizons of the analysis vi) additionally, the reimbursement models within the different counties adoptions of patients choice of provider differed, especially for SLL. 31